Mon 20 Apr 2026 - Balancing homogenous risk groups with credible dataset sizes when selecting a reserving segmentation has long been a challenge for reserving teams. However, if you can get your segmentation right, you can reduce the risk of reserve deteriorations and have a simpler, more efficient reserving process.
During this talk we will present the findings from our latest research into data-driven approaches to identify the optimal reserving segmentation. Our research takes into account data volume and volatility of each segment, and uses hierarchical clustering across a range of characteristics to identify where segments can be grouped together into homogenous groups. A wide range of characteristics can be included in the clustering analysis, including profitability, claims development speed, frequency and severity.
In addition, the optimal reserving segmentation can be overlaid with an analysis of portfolio mix changes to identify where specific changes to the existing reserving segmentation are needed to ensure reserving continues to be appropriate in the future. In the session we will present the methodology and go through case studies of applying the methodology to real-world datasets, for example, the US regulatory Schedule P data.
Ed Harrison and Isabelle Williams, LCP